Radiation exposures from tritium make up a substantial fraction of the occupational and accidental radiation exposures associated with the nuclear power industry. Tritiated water, the most abundant form of tritium, is of particular interest because it is readily taken up by human cells and its irradiation of the cells is spread over a period of days. To approximate the prolonged exposure and the conditions that the cells of an individual would experience in vivo, we irradiated human lymphocytes with tritiated water for 48 h in a 1:1 blood:medium mix. For estimation of the tritium beta-ray dose, a cellular water content of 0.78, based on measurements of human lymphoblastoid cells in culture medium, was used. A modified dose calculation formula was developed for the radiation exposure conditions. A total of 48,014 metaphases (14,482 in irradiated samples and 33,532 in control, unirradiated samples) in human lymphocytes cultured for 72 h after exposure were analyzed for chromosome translocations using fluorescence in situ hybridization. The linear slope (alpha coefficient) of the dose-response curve was measured to be (3.93+/-0.42) x 10(-2) and (5.26+/-0.48) x 10(-2) translocations per cell per gray for complete translocations (tc) and complete translocations plus incomplete translocations [ti(Ab)], respectively, when the data were fitted to a linear model using a weighted least-squares method. The alpha coefficient for tc is significantly lower than that for conventionally measured dicentrics after tritium beta irradiation, but the alpha coefficient for tc + ti(Ab) does not differ significantly from that for dicentrics. This is in agreement with theoretical considerations. The importance of scoring criteria is stressed. The frequency of tc + ti(Ab) is proposed to be a reliable biodosimeter for tritium exposures, and its practical use in a dose reconstruction is presented.